Repository navigation
π Activity monitor: one merge-gate vocabulary and a labelled leading glyph on every gated row - #15084
Conversation
β¦glyph on every gated row At 15:11 ET two Merge gate rows both read 'Ready for auto merge ⬑' while one carried a spinning glyph and the other none. The glyph was the agent-working cue (an agent's lastAction on that PR this round β the review swarm was re-reviewing #15042; nobody was on #15054), unlabelled, so it read as merge progress; and STAGE_TRACKER_ACTIVE_WORD mapped 060, 080 and 100 all to 'Ready for auto merge', so a 060 PR with CI still pending read as ready. - /api/stage-tracker PR items (and nested linked_prs) carry merge_gate {state, reason, sweep_running, last_sweep_at, next_sweep_at, soak_until, soak_remaining_seconds}; automerge gains a SweepClock that records the running pass (cause, targets) and the next floor time. - Pill words in pipeline order: Awaiting merge gate (what holds it) β Ready for auto merge (next sweep in Ns) β Soaking (time left) β Merging β Merged. Every gated row gets a leading glyph with the same sentence as title + aria-label; it spins while a sweep is on the PR, shows a clock while it waits; countdowns tick from the existing cue timer. - The agent-working spinner now carries 'Working now: <agent> β¦' as title + aria-label. - Legend in docs/dashboard-sections.md; node fixture tests on the captured #15042 / #15054 / #15044 payloads; Go tests for merge_gate. Fixes #15073 Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Signed-off-by: Andrew Anderson <andy@clubanderson.com>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
|
@clubanderson is auto-promoted for contributor-lane merging on this PR: the last 5 merged PRs by this author in this hive's repositories have no revert or sentinel-alert evidence. CI, reviewer verdict, diff fences, soak, holds and branch protection still apply. |
|
@clubanderson this PR is held at Failing checks:
Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path. |
Signed-off-by: Andrew Anderson <andy@clubanderson.com>
|
merged by Hive trusted-author auto-merge: author @clubanderson holds role owner and GitHub permission admin on repo; CI green at b7285dc. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
There was a problem hiding this comment.
Decision needed: @clubanderson should decide whether #15084 or #15081 is the fix for #15073 that lands β both add a merge_gate object with the same state vocabulary to the same tracker rows and both add an automerge sweep clock, with different wiring, and they cannot both merge.
This PR and #15081 are competing fixes for the same issue, not duplicates. Both say Fixes #15073, both add MergeGate *StageTrackerMergeGate with json:"merge_gate" to StageTrackerItem and StageTrackerLinkedPR (src/pkg/dashboard/stage_tracker.go), both create src/pkg/dashboard/stage_tracker_merge_gate.go and src/pkg/github/automerge/sweep_clock.go, both use the states gate-pending / ready / soaking / merging / merged, and both instrument automerge_sweep.go and automerge_forgestate.go. I read both diffs in full; the mechanisms differ enough that whichever merges second will conflict on every one of those files:
| #15084 (this PR) | #15081 | |
|---|---|---|
| clock owner | automerge.SweepClock package-level defaultSweepClock, read by the dashboard via automerge.SweepSnapshot() (stageTrackerSweepState) |
github.MergeSweepClock hung off the GitHub client (client.go, merge_sweep_clock.go), reached by the engine through a transport type-assertion (sweepClock()) and by cmd/hive/main.go for the queued-automerge sweep |
| what "running" means | the lane sweep's cause + targeted PR set (SweepState.OnPR; a floor sweep with nil targets marks every gated row) |
per-PR Working(repo, number) marks around the three merge sites, plus Begin("lanes") / Begin("queued") |
| state source | stagelabels.Stage constants + stage record (rec.Reason, soak remaining) |
numeric stage (60/80/95/100/110) + item fields, applied in a post-pass applyMergeGates |
| SPA | 163-line vocabulary rewrite incl. aria-label on the leading glyph |
37-line change |
| extra | api-reference.md citation refresh |
wires the cmd/hive queued sweep too |
Neither one is wrong on its face; this is a pick between two designs (global clock vs client-owned clock; lane-sweep-scoped "working" vs per-PR marks). #15081 already carries a routed verdict at its head, so I am raising the question here rather than there. Both PRs are also labelled conflicts-with-open-pr, which is this.
I have not closed or approved anything.
β hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
The merge-gate pill (#15084) read an issue row's lane reason β 'open PR #15090 covers it (PR at 060-reviewed)' β as a gate reason and drew it as the pill's detail, a two-line clipped label on every issue whose PR sat at the gate. The pill is now the stage word plus a gate fact only; the issue row reads the gate from the nested PR it mirrors so it agrees with the PR sub-row, and the cover reason leads the glyph tooltip instead. Fixes #15101 Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Fixes #15073
What the operator saw
At 15:11 ET on hive's hive, two Merge gate rows both read Ready for auto merge ⬑ β #15042 (base v6) with a leading spinning glyph, #15054 (base v5) without.
Why they differed
The glyph was not merge progress. It is the working cue from #15014 (
.cue-workingβ.ham-active-ico): drawn when some agent'slastActionthis round names the PR. At 15:10:37 the review swarm was re-reviewing #15042 (050β070β050β060 bounce); nobody was on #15054. It had no title oraria-label, so it read as a status indicator. Worse, both rows were at 060 reviewed with CI still pending βSTAGE_TRACKER_ACTIVE_WORDmapped 060, 080 and 100 all to "Ready for auto merge", so the word was wrong too. Nothing to do with v5 vs v6, GitHub auto-merge, or the merge queue.Fix
Payload β
/api/stage-trackerPR items (and the nestedlinked_prscopies) carrybacked by a new
automerge.SweepClockthat records the running pass (cause + targeted PRs) and the next floor pass; both the event-driven loop and the legacy ticker arm it.SPA β one vocabulary in pipeline order, each with its live detail after a
Β·:Every gated row gets a leading glyph with the same plain-words sentence as
titleandaria-labelβ spinning while hive's sweep is on the PR ("hive's merge sweep is running on this PR now."), a clock while it waits ("Approved and green; waiting for hive's merge sweep β next sweep in 42s. A GitHub event on the PR β¦ triggers a pass sooner. Last checked 10s ago."). The agent-working spinner now carries "Working now: β¦" as title + aria-label, so it can never again be mistaken for merge state. Countdowns ride the existing cue tick (no new interval);esc()on every attribute; no inline styles.Actor icon unchanged (existing ⬑ / π€ / π€ per
STAGE_TRACKER_ACTOR); thestage_actorwork in the other PR is not duplicated.Docs β legend table in
src/docs/dashboard-sections.md(the (?) mark stays one sentence per #10915 and links there);labels-and-control-signals.mdwording updated.Tests β
stage_tracker_merge_gate_test.go: Go tests ofstageTrackerMergeGate(all states, targeted vs full pass, soak ageing, off-gate) and ofbuildStageTrackermirroring ontolinked_prs; a node fixture test fed via stdin with the captured #15042 / #15054 / #15044 payloads asserting glyph presence on every gated row, identical title/aria-label, words, details, countdown ticking, andSTAGE_TRACKER_ACTIVE_WORDβ‘STAGE_TRACKER_MERGE_GATE_WORD;sweep_clock_test.goin automerge. Existing row tests updated to the new words.